Parameter Optimization In Image Registration And Fusion Based On Improved Cuckoo Search Algorithm

Image registration attributes seeking optimal spatial transformation to multi-parameter optimization problem. The optimized weighting coefficient in image fusion makes better fusion effect. Fast, accurate and adaptable optimization algorithm is an important step towards the optimization of the parameters. Cuckoo Search(CS) algorithm is a new search algorithm. It derives from the obligate brood parasitic behavior of cuckoos. It has the advantages of few parameter, fast convergence and high efficiency, but there are still existing problems such as search speed is not fast enough, low precision calculation and so on.Based on the framework of CS, the genetic algorithm is introduced and the new algorithm, the Cuckoo Search based on genetic algorithm(GCS) is proposed. GCS adds selecting operation and crossover mutation factor to increase the diversity of population. Experiments results show that the proposed algorithm increased the convergence speed and accuracy. The GCS optimization accuracy is higher, showing better performance of convergence and stability. Another new algorithm, the Cuckoo Search based on chaos search(CHCS) is also proposed. It takes advantage of chaotic motion randomness, ergodicity features to make the population evenly distributed in space and search for optimal solutions to avoid falling into local extreme. Experiments results show that algorithm has some performance improvements for low-dimensional function optimization, the convergence performance and stability are poor than GCS.The parameters of standard CS are kept constant, which may affect the convergence and accuracy of the algorithm. In order to improve the algorithm's speed and calculation accuracy, an adaptive genetic factors Cuckoo Search(AGCS) is proposed, which the step length factor and find probability change in the iterations. Experiments show that the AGCS has better convergence accuracy and convergence speed. Finally it is used for parameter optimization of image registration and fusion in the paper. The experimental results show that the AGCS algorithm makes registration time faster and registration precision higher. And it finds that the optimal weight coefficient makes the fused images show more comprehensive and useful information. This algorithm in this paper is better than other traditional algorithm so it is effective, stable and feasible.